Organizational Memory: Capturing Context Before It Walks Out the Door

The CIO of NASA told me about a brilliant idea that combines the contextual knowledge of subject-matter experts with the processing power of AI. 

We were talking about knowledge management and what happens when experienced people leave an organization. NASA, like many large organizations, has employees who may spend decades working on highly specialized missions, systems and processes. When they retire or move on, much of their learned knowledge leaves with them. 

To counter this loss of knowledge and organizational context, they started conducting structured interviews with people before they left. They would ask a defined set of questions, record the conversations and capture that knowledge so it could be referenced using a large language model. Obviously, this wasn’t a replacement for the actual person, but in the context of losing all that experience, it was better than nothing. 

I often think back to that discussion, in part because I thought it was a brilliant idea, especially given that this was a couple of years ago, and also because it highlights something many organizations struggle with: you don’t know what you don’t know, even when it’s gone. The loss of institutional context can be especially difficult because you may not realize what disappeared until you encounter a situation where you need it. Unlike a missing document or file, the loss itself can be invisible. By the time you recognize the gap, the window to recover that context may already have closed. 

The Knowledge That Never Makes It Into the Process 

Most organizations are very good at documenting certain kinds of data. We have policies, procedures, project plans and email archives. There is no shortage of data. But data and organizational knowledge are not the same thing. 

Think about an experienced employee who has worked within the same program for 15 years. They know which part of the official process matters most, which stakeholders need to be brought into a decision early, which exceptions occur regularly, and which ones should immediately raise concern. They may remember why a process was designed a certain way even though the original decision was never formally documented. They also know the workarounds. The document says one thing, but the people doing the work know there are three additional steps required to actually get it done. 

That is context, and a surprising amount of it exists only in people’s heads. 

AI Is Making the Blind Spot More Visible 

None of this is new. Organizations have been losing institutional knowledge through transitions and attrition forever. What AI is doing is making those gaps in context much more visible. 

When organizations begin exploring AI, one of the first questions is: What data does the AI need to work for this use case? That quickly leads to other questions: 

• Where does that information live? 
• Which source is authoritative? 
• Why was this rule created? 
• What happens when there is an exception? 
• Who makes the decision in that situation? 

And, all too often, the initiative stops at this point because it forces the organization to understand itself at a depth that simply isn’t documented. 

We Need to Capture the Why, Not Just the What 

Traditionally, we have focused heavily on preserving artifacts: put everything in SharePoint, make sure all tickets go in Jira. Those are helpful for organizing and preserving information, but they often capture what happened without capturing why it happened. 

Imagine finding a five-year-old project plan that shows a major change halfway through implementation. You can see the change, but can you determine why the decision was made? In an instance like this, the context may be more valuable than the document itself. 

Where AI Can Actually Help 

We still need to preserve organizational artifacts. Repositories like SharePoint and Jira are treasure troves of data. They are necessary, but they are not sufficient. 

The NASA example is so compelling because it uses AI to capture context in a way that would have been much harder before. The structured interview creates a deliberate way of understanding what someone knows, preserving it, and then making that knowledge available for future use. 

I continue to believe that the purpose of technology, including AI, is to empower people. In this case, that means preserving the context and experience people have built over years and making it available to others when they need it. AI can help make that knowledge easier to find and use, while people remain responsible for deciding what is relevant, current and appropriate to apply. 

Preserving Context Before It’s Needed 

The challenge, of course, is that organizations often recognize the value of this knowledge only when someone is about to leave. By then, the window to capture years of experience may be too small. The bigger opportunity is to capture and preserve the right context as the work and decisions are happening, rather than waiting for a transition to force the issue. 

We are never going to capture everything a person knows, and that should not be the goal. But there is value in being more intentional about preserving the context and the why behind how work gets done, especially in areas where a small number of people hold a disproportionate amount of institutional knowledge. 

How you do this is up to you. But I can think of a lot worse leads to follow than NASA.